This paper is intended to support the preservation of national cultural asset, particularly for ancient symbols. By using image processing principle, an automatic system that can be designed and implemented to translate ancient manuscript documents. The system is composed of several phases, from scanning, preprocessing, segmentation, feature extraction and classification. Sample images of the document are not scanned automatically, but manually produced as monochrome, black for the text and white for the background. These sample images are varied based on font size, rotation, and image size. The system is intended to be adaptable for various condition except for the color variation. The system is implemented as a MATLAB application program to convert an image that contains random Batak symbols into a series of Latin character representation of each word. The experiment results show that the system accuracy is ranged between 42% - 96% and the processing time is ranged from 1.9 - 34 seconds. © 2015 IEEE.